Let-7 represses Nr6a1 and a mid-gestation developmental program in adult fibroblasts

Abstract

MicroRNAs (miRNAs) are critical to proliferation, differentiation, and development. Here, we characterize gene expression in murine Dicer-null adult mesenchymal stem cell lines, a fibroblast cell type. Loss of Dicer leads to derepression of let-7 targets at levels that exceed 10-fold to 100-fold with increases in transcription. Direct and indirect targets of this miRNA belong to a mid-gestation embryonic program that encompasses known oncofetal genes as well as oncogenes not previously associated with an embryonic state. Surprisingly, this mid-gestation program represents a distinct period that occurs between the pluripotent state of the inner cell mass at embryonic day 3.5 (E3.5) and the induction of let-7 upon differentiation at E10.5. Within this mid-gestation program, we characterize the let-7 target Nr6a1, an embryonic transcriptional repressor that regulates gene expression in adult fibroblasts following miRNA loss. In total, let-7 is required for the continual suppression of embryonic gene expression in adult cells, a mechanism that may underlie its tumor-suppressive function.

Figure 1.

Loss of Dicer in adult MSCs results in specific gene expression changes. (A) Waterfall plot of genes expressed differentially by exon microarray at an adjusted P-value ≤0.1. Two-hundred-seventeen genes are shown. Dashed gray lines indicate median up-regulation (threefold) and down-regulation (threefold). Two Dicer1^f/f (Dicer wild-type [WT]) and four Dicer1^−/− (Dicer knockout [KO]) clones were evaluated for gene expression changes. (B) Scatter plot of the median change in expression of miRNA targets relative to control gene sets matched for 3′ UTR length, GC content, and expression in Dicer wild-type MSCs. Each point represents conserved targets of a single miRNA seed family. miRNA expression is based on the small RNA-seq data reported in this study. The P-value, whose negative log₁₀ is shown on the X-axis, was calculated by Wilcoxon rank-sum test. (C) Cross-correlation analysis of Dicer wild-type and knockout MSCs versus a cell line panel based on total gene expression by exon microarray. The gene expression of Dicer wild-type and knockout ESCs was published previously (Leung et al. 2011). The gene expression of all other murine cell lines is publicly available from Novartis BioGPS (Wu et al. 2009, 2013). See also Supplemental Figure S1 and Supplemental Tables S1 and S2.

Figure 2.

Dicer knockout (KO) MSCs up-regulate oncofetal genes. (A,B) qPCR (A) and Western blot analysis (B) of oncofetal genes. p107 is shown as a loading control for Western blot. Error bars indicate standard error of the mean (±SEM). (C) Normalized read counts for chromatin marks at Igf2bp1 for Dicer wild-type (WT) and knockout MSCs. Two replicates (r1 and r2) are shown per sample. Within each chromatin mark (H3K4me3, H3K36me3, and H3K27me3), all samples are set to the same scale. Flanking genes are shown as controls. (D) Box plot of log₂ fold change in gene expression for all genes (“All”), all predicted let-7 targets (“Let-7”), genes enriched in H3K4me3 in Dicer knockout MSCs relative to Dicer wild-type MSCs (“H3K4me3 in Dicer KO”), or overlapping genes from the latter two categories (“Let-7, H3K4me3 in Dicer KO”). P-values were calculated with a Wilcoxon rank-sum test. See also Supplemental Figure S2 and Supplemental Tables S2 and S3.

Figure 3.

Identification of let-7 targets by let-7 add-back to Dicer knockout (KO) MSCs. (A) Heat map of expression Z-scores in transfected Dicer wild-type (WT) and knockout MSCs for all genes expressed at FPKM ≥0.1. Genes are ranked from highest (top) to lowest (bottom) Z-score in Dicer knockout MSCs. A 10-gene moving average of TargetScan scores is shown in the left plot for let-7 (black) and miR-15 (gray). (B) Overlap of genes up-regulated at q-value <0.05 in siCtrl-transfected Dicer knockout MSCs relative to siCtrl-transfected Dicer wild-type MSCs (“Derepressed in Dicer KO”), down-regulated at q-value <0.05 in let-7g-transfected Dicer knockout MSCs relative to siCtrl-transfected Dicer knockout MSCs (“Repressed with let-7 add-back in Dicer KO”), and predicted by TargetScan to be conserved targets of let-7 (“Predicted targets of let-7”). Top predicted let-7 targets in TargetScanMouse are shown in the table on the right. (C) qPCR analysis of high-confidence targets of let-7 identified in triple overlap above. Error bars indicate the SEM. (D) Metaplot of CLIP-seq read density. CLIP-seq reads were aligned to TargetScan-predicted sites. The average CLIP-seq read coverage per gene was plotted relative to let-7 target sites within the 3′ UTRs of high-confidence targets of let-7 (“HC, let-7”), let-7 target sites within the 3′ UTRs of all TargetScan-predicted targets of let-7 (“All, let-7”), or miR-124 target sites within the 3′ UTRs of all TargetScan-predicted targets of let-7 (“All, miR-124”). The analysis was carried out for genes with FPKM ≥0.1 in either Dicer wild-type or knockout MSCs transfected with siCtrl. (E) CLIP-seq density along the 3′ UTR of Nr6a1. The blue histogram indicates CLIP-seq density; gray tick marks indicate the location of TargetScan-predicted, conserved let-7 sites; and orange tick marks indicate the location of TargetScan-predicted, nonconserved let-7 sites. The length of the 3′ UTR is indicated. See also Supplemental Figures S3 and S4 and Supplemental Tables S4 and S6.

Figure 4.

Let-7 targets are dynamically expressed in the whole mouse embryo. (A) Heat map of individual high-confidence let-7 targets in the whole mouse embryo from E7.5 to E18.5 (Irie and Kuratani 2011). Genes validated by qPCR in this study are labeled on the right. (B–E) Average expression Z-score for high-confidence let-7 targets (B), all predicted conserved TargetScan targets of let-7 (C), indirect targets of let-7 (D), and Nr6a1 (E) in the whole mouse embryo from E7.5–E18.5 (Irie and Kuratani 2011). The black dashed line shows let-7 expression. Error bars indicate the SEM. See also Supplemental Figure S5.

Figure 5.

Gain-of-function and loss-of-function approaches identify genes regulated by let-7 target NR6A1. (A) Schematic of experimental design. (From left) Dicer wild-type (WT) MSCs (“NR6A1-low”) were infected with pMMP-puro retrovirus encoding vector control or Flag-HA-NR6A1, selected with puromycin, and passaged for several weeks prior to isolation of total RNA for polyA-selected mRNA-seq. (From right) Dicer knockout (KO) MSCs (“NR6A1-high”) were transfected with control nontargeting siRNA (siCtrl) or siRNA against Nr6a1 (siNR6A1) 48 h prior to isolation of total RNA for polyA-selected mRNA-seq. (B) Box plot of gene expression changes in siCtrl-transfected Dicer knockout versus siCtrl-transfected Dicer wild-type MSCs for all genes (“All genes”) or genes differentially expressed upon overexpression of Flag-HA-NR6A1 (“Flag-HA-NR6A1-responsive”). An FPKM cutoff ≥0.1 was used for all compared data sets. P-value was calculated by Wilcoxon rank-sum test. (C) Box plot of gene expression changes in siNR6A1-transfected Dicer knockout versus siCtrl-transfected Dicer knockout MSCs for all genes (“All genes”) or genes differentially expressed upon overexpression of Flag-HA-NR6A1 (“Flag-HA-NR6A1-responsive”). An FPKM cutoff ≥0.1 was used for all compared data sets. P-value was calculated by Wilcoxon rank-sum test. (D) Triple overlap of Flag-HA-NR6A1-responsive, NR6A1 knockdown-responsive, and Dicer deletion-responsive gene sets. (E) Average Z-score in the whole mouse embryo time course for 32 NR6A1-responsive genes defined in D. The orange dashed line shows Nr6a1 expression, and the black dashed line shows let-7 expression. Error bars indicate the SEM. See also Supplemental Figure S6 and Supplemental Table S4.

Figure 6.

Global genomic profile of NR6A1 binding. (A) Summary of binding data for Flag-HA-NR6A1. In total, 9223 sites are bound genome-wide and map to either promoter-proximal (5 kb upstream of to 1 kb downstream from the TSS), genic (excluding promoter-proximal regions), or intergenic regions. Associated genes were defined as those genes proximal to promoter peaks, overlapping with genic peaks, or closest to intergenic peaks. In total, 5210 genes are associated with or near Flag-HA-NR6A1-binding sites. The called regions represent the intersection of two clonal replicates. (B) Motifs identified by DREME analysis in each category (promoter, genic, and intergenic) summarized in A. The motifs are ranked by enrichment score (E-value) within each category (rank in category). The published motif is also shown (Yan et al. 1997). (C) Normalized read counts of two clonal replicates (r1 and r2) of vector control or Flag-HA-NR6A1 at Nmnat3, a gene responsive to Flag-HA-NR6A1 overexpression, NR6A1 knockdown, and Dicer loss. The promoter-proximal site includes two tandem NR6A1 consensus motifs (shown above the read counts). A flanking gene is shown as a control. (D,E) Box plots of gene expression changes in Flag-HA-NR6A1-overexpressing Dicer wild-type (WT) MSCs versus vector-only Dicer wild-type MSCs (D) or siNR6A1-transfected Dicer knockout (KO) versus siCtrl-transfected Dicer knockout MSCs (E) for all genes (“All genes”), the 32 genes responsive to NR6A1 as defined by triple overlap (“Responsive”), genes responsive to and bound by NR6A1 (“Responsive, bound”), or the subset of NR6A1 consensus motif-containing genes responsive to and bound by NR6A1 (“Responsive, bound, motif”). See also Supplemental Figure S7 and Supplemental Tables S3 and S4.

Figure 7.

Summary of let-7 and target expression in the whole mouse embryo. Let-7 targets peak mid-gestation (around E8.5–E10.5), after down-regulation of the ESC-specific miR-290 family. Expression of mature let-7 becomes detectable in the whole embryo around E10.5 and steadily increases in level, concomitant with down-regulation of high-confidence let-7 targets such as Nr6a1, identified in this study. Targets of Nr6a1 identified in this study increase as the embryo matures and positively correlate with let-7 in the whole mouse embryo.